diff --git a/scrapy/http/response/html.py b/scrapy/http/response/html.py
index dc812ac0e..1ef52490c 100644
--- a/scrapy/http/response/html.py
+++ b/scrapy/http/response/html.py
@@ -18,15 +18,15 @@ class HtmlResponse(TextResponse):
 
     _httpequiv_re = _template % ('http-equiv', 'Content-Type')
     _content_re   = _template % ('content', r'(?P<mime>[^;]+);\s*charset=(?P<charset>[\w-]+)')
-    _encoding_re  = _template % ('encoding', r'(?P<charset>[\w-]+)')
+    _content2_re   = _template % ('charset', r'(?P<charset>[\w-]+)')
 
     METATAG_RE  = re.compile(r'<meta\s+%s\s+%s' % (_httpequiv_re, _content_re), re.I)
-    METATAG_RE2 = re.compile(r'<meta\s+%s\s+%s' % (_content_re, _httpequiv_re), re.I)
+    METATAG2_RE  = re.compile(r'<meta\s+%s' % _content2_re, re.I)
 
     @memoizemethod_noargs
     def _body_declared_encoding(self):
         chunk = self.body[:5000]
-        match = self.METATAG_RE.search(chunk) or self.METATAG_RE2.search(chunk)
+        match = self.METATAG_RE.search(chunk) or self.METATAG2_RE.search(chunk)
         return match.group('charset') if match else None
 
 
diff --git a/scrapy/tests/test_http_response.py b/scrapy/tests/test_http_response.py
index 3c6afdcf9..b1edf4a3a 100644
--- a/scrapy/tests/test_http_response.py
+++ b/scrapy/tests/test_http_response.py
@@ -260,6 +260,10 @@ class HtmlResponseTest(TextResponseTest):
         r4 = r3.replace(body=body)
         self._assert_response_values(r4, 'iso-8859-1', body)
 
+    def test_html5_meta_charset(self):
+        body = """<html><head><meta charset="gb2312" /><title>Some page</title><body>bla bla</body>"""
+        r1 = self.response_class("http://www.example.com", body=body)
+        self._assert_response_values(r1, 'gb2312', body)
 
 
 class XmlResponseTest(TextResponseTest):
